Classic DNS was designed in the 1980s on a simple assumption: if you ask a question and an answer comes back, it's probably true. That assumption has aged badly. Anyone able to slip a forged reply in front of the real one can send visitors to the wrong server, and the visitor's browser will never know.

DNSSEC is the fix. It doesn't hide anything, and it doesn't make DNS faster. What it does is add a cryptographic signature to answers so a resolver can check they really came from the domain's owner and haven't been altered on the way.

The problem it solves

The best-known attack is cache poisoning. In 2008 researcher Dan Kaminsky showed that a resolver could be tricked into storing a forged answer, which would then be served to every user of that resolver until it expired. Randomising ports and query IDs made the attack much harder, but not impossible, and nothing in plain DNS lets a resolver prove an answer is genuine. DNSSEC supplies that proof.

How the signatures chain together

DNSSEC uses a handful of extra record types, and they fit together like links in a chain.

  • RRSIG is the signature itself, attached to each set of records.
  • DNSKEY is the public key a resolver uses to check those signatures. It's published in the zone it protects.
  • DS (delegation signer) is a fingerprint of that key, stored in the parent zone. This is the crucial link.

A validating resolver starts at the root, whose key it already trusts. The root vouches for .com through a DS record, .com vouches for your domain the same way, and your domain's key then signs your actual records. If any link is missing or wrong, the whole answer is treated as untrustworthy.

We looked at a real, signed domain. Asking Cloudflare's resolver for cloudflare.com with DNSSEC data requested returned this:

;; flags: qr rd ra ad; QUERY: 1, ANSWER: 3
cloudflare.com.  300  IN  A      104.16.132.229
cloudflare.com.  300  IN  A      104.16.133.229
cloudflare.com.  300  IN  RRSIG  A 13 2 300 20261123122712 ...

Two details matter. The ad flag (authenticated data) means the resolver checked the signature chain and was satisfied. And the 13 in the RRSIG is the algorithm number, ECDSA with the P-256 curve, which is the modern default because its signatures are small. The DS record sitting at the parent for that domain had a TTL of 86400, one day.

What DNSSEC does not do

It's easy to oversell, so here are the limits.

  • No encryption. Anyone on the network can still see which names you look up. Encrypted transports like DNS over HTTPS cover that, and they're a separate thing.
  • No protection against look-alike domains. A phishing site at examp1e.com can be perfectly signed.
  • No protection for the last hop. If your laptop doesn't validate, or trusts a resolver that doesn't, the signatures are never checked.
  • It only helps if people validate. Major public resolvers do, which is why signing your domain now has real effect.

The downside nobody mentions in the brochure

A misconfigured DNSSEC setup doesn't degrade gracefully. It takes your domain offline for everyone whose resolver validates. A validating resolver that can't verify an answer returns SERVFAIL, and users simply see an error.

The failures that cause this are surprisingly boring:

  1. Moving DNS providers without handling the DS record. The registry still advertises the old key, the new provider signs with a different one, and nothing matches. This is the most common cause by far.
  2. Expired signatures. RRSIGs have an expiry date. If your provider's automation stops re-signing, answers die on schedule.
  3. A botched key rollover when keys are replaced by hand.

How to tell whether DNSSEC is the culprit

There's a trick that saves a lot of head-scratching. Ask the same question twice, once normally and once with the "checking disabled" flag:

dig example.com A @1.1.1.1
dig example.com A @1.1.1.1 +cd

If the first command returns SERVFAIL and the second returns a normal answer, the data is fine and the signatures are broken. For a closer look, dig +dnssec shows the signatures, and delv (shipped with BIND tools) performs the validation itself and tells you which link failed.

Does DNSSEC slow things down?

Slightly in theory, and rarely in a way anyone can notice. Signed answers are larger because they carry signatures, which occasionally pushes a response past what fits in one UDP packet and makes the resolver retry over TCP. The validation work happens at the resolver, which caches the verified result, so the cost is paid once for many visitors. The real risk with DNSSEC is misconfiguration, not performance.

Should you turn it on?

Our view: yes, if your DNS provider automates it. When a provider signs the zone and gives you a ready-made DS record to paste at your registrar, the risk is small and the protection is real. Managing keys yourself is only worth it if you have a reason and a monitoring setup to match.

The safe order of operations is the same every time:

  1. Enable signing at the DNS provider.
  2. Wait until the signed zone is serving everywhere.
  3. Add the DS record at the registrar.
  4. Check with a validating resolver.

When you move providers, reverse it: remove the DS record first, wait out its TTL (often a day), move the zone, then re-enable DNSSEC at the new home. Our guide on changing nameservers without downtime includes this in its checklist.

Keeping an eye on it

Run an NS and DS check from several places after any change, using the DNS checker. If some resolvers answer and others return an error, suspect signatures before you suspect propagation, and compare it with what our article on DNS propagation says about normal delays.